HPV-Stratified Causal Digital Twins for Personalized Survival Benefit Estimation in Head and Neck Cancer
- 1Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA
Abstract
PurposeDigital-twin frameworks can potentially offer a pathway toward individualized treatment but remain largely untested in radiation oncology. We present a causal digital-twin model for head-and-neck (H&N) cancer that estimates patient-specific survival benefit from chemoradiotherapy (CRT) versus radiotherapy (RT) alone, stratified by HPV status.
MethodsUsing the public RADCURE cohort (n = 3,346), we developed a causal digital-twin framework in which patients were divided into HPV-positive and HPV-negative groups and assigned to training (RT start year [≤] 2007) and test sets (RT start year > 2007). Bayesian-network structure learning identified each cohorts Markov Blanket around the 2-year survival endpoint (SVy2), yielding the minimal causally relevant feature set. A causal-forest model (double machine learning, DML framework) estimated the individualized treatment effect{tau} = P(Svy2 | CRT,X) - P(Svy2 | RT,X), with [Formula] denoting the model-estimated effect. Overlap restrictions were applied to mitigate extrapolation. Predictive discrimination (AUC, C-index), average treatment effects (ATE), individualized effects (ITE), and decision-curve analyses were computed.
ResultsFor HPV-negative disease, internal evaluation demonstrated strong discriminative performance (test AUC = 0.72; C-index = 0.81, p < 0.01). The estimated test-set ATE favored CRT with a modest but positive improvement in SVy2 (mean [Formula]). Patterns in [Formula] were clinically coherent, including increased benefit in node-positive and good-performance patients and reduced benefit in high-burden or poor-performance presentations. Decision-curve analysis showed that a [Formula]-guided selective-CRT strategy provided higher net benefit than uniform treatment policies. HPV-positive disease exhibited substantially greater but less clinically structured treatment-effect heterogeneity, accompanied by reduced overlap and greater uncertainty.
ConclusionThis study demonstrates a proof-of-concept causal digital-twin framework for interpretable, patient-specific survival simulation in H&N cancer. The model captures clinically plausible treatment-effect heterogeneity and establishes a foundation for multi-cohort validation and future integration into individualized treatment decision support.
Related
- No Overall Survival Benefit with Adding Chemotherapy to Immunotherapy in PD-L1 TPS >= 50% NSCLC: An Agent-Stratified Reassessment
- Examined lymph node count and survival in resected non-small cell lung cancer: attenuation of benefit beyond 16 nodes is confined to node-negative disease
- Number of examined lymph nodes and overall survival in intrahepatic cholangiocarcinoma: a population-based study suggesting a staging benefit
- Cysteine reduction rather than methionine restriction promotes immunometabolic fitness to confer survival benefit in aging
- Apparent Survival Benefit of Postoperative Radioactive Iodine in Lymph Node-Positive Papillary Thyroid Carcinoma: A Negative-Control Analysis for Unmeasured Confounding.
BibTeX
@misc{shah2026hpvstratified,
title = {{HPV-Stratified Causal Digital Twins for Personalized Survival Benefit Estimation in Head and Neck Cancer}},
author = {K. D. Shah and C.-W. Chang and J. Zhou and J. Ding and A. Dhabaan and H. A. Al-Hallaq and J. Bates and W. Stokes and D. S. Yu and X. Yang},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.30.26363910},
url = {https://www.medrxiv.org/content/10.64898/2026.09.30.26363910v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - HPV-Stratified Causal Digital Twins for Personalized Survival Benefit Estimation in Head and Neck Cancer
AU - K. D. Shah
AU - C.-W. Chang
AU - J. Zhou
AU - J. Ding
AU - A. Dhabaan
AU - H. A. Al-Hallaq
AU - J. Bates
AU - W. Stokes
AU - D. S. Yu
AU - X. Yang
PY - 2026
DA - 2026/09/30
PB - medRxiv
DO - 10.64898/2026.09.30.26363910
UR - https://www.medrxiv.org/content/10.64898/2026.09.30.26363910v1
AB - PurposeDigital-twin frameworks can potentially offer a pathway toward individualized treatment but remain largely untested in radiation oncology. We present a causal digital-twin model for head-and-neck (H&N) cancer that estimates patient-specific survival benefit from chemoradiotherapy (CRT) versus radiotherapy (RT) alone, stratified by HPV status. MethodsUsing the public RADCURE cohort (n = 3,346), we developed a causal digital-twin framework in which patients were divided into HPV-positive and HPV-negative groups and assigned to training (RT start year [≤] 2007) and test sets (RT start year > 2007). Bayesian-network structure learning identified each cohorts Markov Blanket around the 2-year survival endpoint (SVy2), yielding the minimal causally relevant feature set. A causal-forest model (double machine learning, DML framework) estimated the individualized treatment effect{tau} = P(Svy2 | CRT,X) - P(Svy2 | RT,X), with [Formula] denoting the model-estimated effect. Overlap restrictions were applied to mitigate extrapolation. Predictive discrimination (AUC, C-index), average treatment effects (ATE), individualized effects (ITE), and decision-curve analyses were computed. ResultsFor HPV-negative disease, internal evaluation demonstrated strong discriminative performance (test AUC = 0.72; C-index = 0.81, p < 0.01). The estimated test-set ATE favored CRT with a modest but positive improvement in SVy2 (mean [Formula]). Patterns in [Formula] were clinically coherent, including increased benefit in node-positive and good-performance patients and reduced benefit in high-burden or poor-performance presentations. Decision-curve analysis showed that a [Formula]-guided selective-CRT strategy provided higher net benefit than uniform treatment policies. HPV-positive disease exhibited substantially greater but less clinically structured treatment-effect heterogeneity, accompanied by reduced overlap and greater uncertainty. ConclusionThis study demonstrates a proof-of-concept causal digital-twin framework for interpretable, patient-specific survival simulation in H&N cancer. The model captures clinically plausible treatment-effect heterogeneity and establishes a foundation for multi-cohort validation and future integration into individualized treatment decision support.
ER -